Convert from RGB to Grayscale without rgb2gray

Basically, I need to make a function [gray] = grayscale(color) without using rgb2gray. Basically, I am taking in a 3x3 matrix with color and returning a 2x2 matrix with gray. I get the fact that I have to extract the red, green and blue channels using:

redChannel = color(:, :, 1);
greenChannel = color(:, :, 2);
blueChannel = color(:, :, 3);

and then using the weighted average:

gray = .299*redChannel + .587*greenChannel + .114*blueChannel

but I do not know how to make that gray into a 2x2 matrix. Any help is appreciated.

[DGM]

See also:

https://www.mathworks.com/matlabcentral/answers/99136-how-do-i-convert-my-rgb-image-to-grayscale-without-using-the-image-processing-toolbox

and

https://www.mathworks.com/help/matlab/ref/function.html

https://www.mathworks.com/help/matlab/matlab_prog/create-functions-in-files.html

Answer 1

[Image Analyst]

Your 3x3 and 2x2 jargon is all messed up so I will ignore it. You're almost there with your function though. First I wouldn't use color as the name of a variable since it's so often the name of a parameter. So let's call it rgbImage. So your function is

function grayImage = grayscale(rgbImage)
try
  [rows, columns, numberOfColorChannels = size(rgbImage);
  if numberOfColorChannels  == 3
      % It's color, need to convert it to grayscale.
      redChannel = rgbImage(:, :, 1);
      greenChannel = rgbImage(:, :, 2);
      blueChannel = rgbImage(:, :, 3);
      % Do the weighted sum.
      grayImage = .299*double(redChannel) + ...
                  .587*double(greenChannel) + ...
                  .114*double(blueChannel);
      % You probably want to convert it back to uint8 so you can display it.
      grayImage = uint8(grayImage);
  else
      % It's already gray scale.
      grayImage = rgbImage;  % Input image is not really RGB color.
  end
catch ME
  errorMessage = sprintf('Error in function %s() at line %d.\n\nError Message:\n%s', ...
    ME.stack(1).name, ME.stack(1).line, ME.message);
  fprintf(1, '%s\n', errorMessage);
  uiwait(warndlg(errorMessage));
end

This will take a rows-by-columns-by-3 image (a 3-D image) and convert it to a rows-by-columns image (a 2-D image). You could make it a little more robust by checking that numberOfColor channels is either 1 or 3 and throw an error if it's not, but that likely won't happen if you pass it what you're supposed to.

[Walter Roberson]

[rows, columns, numberOfColorChannels = size(rgbImage);

needs to be

[rows, columns, numberOfColorChannels] = size(rgbImage);

[Laura Castro]

Great piece of code

One have to close this bracket here

[rows, columns, numberOfColorChannels]

and add an end -at the end.

... ME.stack(1).name, ME.stack(1).line, ME.message);

fprintf(1, '%s\n', errorMessage);

uiwait(warndlg(errorMessage));

end

end

Other than that, best current solution for this headache.

Nicely done,

Thanks Walter!

[thanh nguyen]

can you help me perform this code in parallel processing? pls

[Walter Roberson]

@thanh nguyen

Why would you need to do this yourself in parallel, instead of calling rgb2gray() inside your parallel code?

Anyhow, if you stored Image Analyst's code into a file, you can invoke the file from your parallel regions. Just remember to make the fix

[rows, columns, numberOfColorChannels] = size(rgbImage);

[thanh nguyen]

i don't understand : "weighted sum" ? what is it? is it the fomula built-in?

[Walter Roberson]

The numbers are from CCIR 601

https://en.wikipedia.org/wiki/Luma_(video)#Rec._601_luma_versus_Rec._709_luma_coefficients